The choice of a point of articulation for a given sound involves the isolation of one region of the vocal tract as fundamental, despite the fact that significant articulatory activity may be taking place elsewhere as well. It is usually assumed that primary and secondary articulation can be distinguished by relative degree of occlusion; this is impossible, however, in multiply-articulated stops ([kp, gb] etc.), spirants, and glides (including [w]). Evidence is reviewed that suggests that even in these cases one or the other occlusion should be regarded as primary. In some languages, “labial-velars” function as labials; in others, as velars. In at least one language, Fula, labial and velar types are in contrast. In none of these cases, however, need there be any phonetic distinction associated with the contrast. A segment’s phonological behavior within the language may be the only indication of its character. From these facts, we conclude that (1) the traditional distinction of primary and secondary articulation is valid even for segments with more than one occlusion of the same degree; and (2) at least some descriptive phonetic parameters cannot be determined directly from an inspection of the phonetic substance, but only from the consideration of alternations and related forms.